-
Notifications
You must be signed in to change notification settings - Fork 55
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Automatically Determine Manifold Features for Testing and Documentation #650
Conversation
Codecov Report
@@ Coverage Diff @@
## master #650 +/- ##
==========================================
- Coverage 99.22% 98.07% -1.15%
==========================================
Files 106 107 +1
Lines 10430 10552 +122
==========================================
Hits 10349 10349
- Misses 81 203 +122
📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more |
What other properties would you put there? I'd imagine you could for example consider |
My idea is to collect a few “features” that might be nice to know in a summary table (that lists functions, retractions, inverse retractions, vectors transports) for a manifold. For example whether it has an embedding or so. But really just things that are known generically. |
Hm, "things that are known generically" sounds very restrictive to me. I'd guess that for everything other than |
manifold dimension and rand would not fit so much; as I said I am not yet 100% sure, my main idea was to collect traits. |
If you just want traits then there won't be much as IMO traits should be limited to what we need to dispatch on. Also, functions::Vector{F}
retractions::Vector{AbstractRetractionMethod}
inverse_retractions::Vector{AbstractInverseRetractionMethod}
vector_transports::Vector{AbstractVectorTransportMethod} doesn't look like collecting traits? |
I was only speaking about the The ones you mention – that is the easy and solved part – I collect functions, retractions, inverse retractions an vector transports that are available on some M. But for automatic testing features like – does it have an embedding and is it isometric would be cool to generically know. That is what I have not yet modelled completely here. |
I was actually commenting that it's only easy when you have fixed parameters for a manifold. For example a table for |
For the tests that is intentional – for the docs and displaying a summary that might be something to still check. |
OK. One more point, when re-running feature discovery it would be good to have something for comparing two saved feature sets to make sure the newly added feature is correctly discovered and no old features are marked as not present. Maybe just saving to a text-based format so that standard diff tools work or, if the format is binary, a function that reports changes. |
That might be details a bit more necessary on the other branch, but I did not plan to design a full text file export, but just save the jd2. One could still to a set-based difference report if something changes – especially if something is removed. |
Concerning the properties, I will aim to do a minimal approach here and keep that flexible for now, it depends a bit on what we can do with that in the tests in the end. |
So for now the features produce something like
And I think I would like to keep it at that for now. The features will be nice for knowing whether the manifold is embedded and in what sense (so one could spare an Still nt yet completly determined here (still to check how to best find that)
|
This seems to work reasonably well, but I will merge this into the other PR, since I can not do that well of a testing here; |
In order to properly do #649, we need a way to determine which functions exist (at a certain time, cf. the discussion at #548) on a manifold. This revises an approach that I did in #548 already. This feature approach had two ideas (testing and documentation) I would like to finish before continuing on the testing.
The main part I am not yet sure about is how to provide the
properties
section – slightly whether that is useful in testing (embedding is, but the rest I am not sure), but mainly which parts one could add for the documentation here in a reasonable way.